ABSTRACT
Human monocytotropic ehrlichiosis (HME), caused by the bacterium Ehrlichia chaffeensis, and human granulocytic anaplasmosis (HGA), caused by the bacterium Anaplasma phagocytophilum, are two emerging tick-borne zoonoses of concern. Factors influencing geographic distributions of these pathogens are not fully understood, especially at varying spatial extents (regional versus landscape) and resolutions (counties versus smaller land units). We used logistic regression to compare influences of physical environment, land cover composition, and landscape heterogeneity on distributions of A. phagocytophilum and E. chaffeensis at multiple spatial extents. Pathogen presence or absence was determined from white-tailed deer (Odocoileus virginianus) serum samples collected from 1981 to 2005. Ecological predictor variables were derived from spatial datasets that represented deer density, elevation, land cover, normalized difference vegetation index (NDVI), hydrology, and soil moisture. We used three strategies (a priori, exploratory, and spatial extent) to develop models. Best fitting models were applied within a geographic information system to create predictive probability surfaces for each bacterium. Ecological predictor variables generally resulted in better fitting models for E. chaffeensis than A. phagocytophilum (90.5% and 68% sensitivity, respectively), possibly as a result of differences in the natural histories of tick vectors. Although alternative model development strategies produced different models, in all cases bacteria presence or absence was affected by a combination of soil moisture or flooding variables (thought to affect primarily tick vectors) and forest cover or NDVI variables (thought to affect primarily mammalian hosts). This research demonstrates the potential for modeling the distributions of microscopic tick-borne pathogens using coarse regional datasets and emphasizes the importance of forest cover and flooding as environmental constraints, as well as the importance of considering ecological variables at multiple spatial extents.
